How should we measure collaborative research and development performance ?